Herschel/HIFI observations of a new interstellar water maser: the 5(32)-4(41) transition at 620.701 GHz

Using the Herschel Space Observatory's Heterodyne Instrument for the Far-Infrared (HIFI), we have performed mapping observations of the 620.701 GHz 5(32)-4(41) transition of ortho-H2O within a roughly 1.5 x 1.5 arcmin region encompassing the Kleinmann-Low nebula in Orion, and pointed observations of that transition toward the Orion South condensation and the W49N region of high-mass star formation. Using the Effelsberg 100 m radio telescope, we obtained ancillary observations of the 22.23508 GHz 6(16)-5(23) water maser transition; in the case of Orion-KL, the 621 GHz and 22 GHz observations were carried out within 10 days of each other. The 621 GHz water line emission shows clear evidence for strong maser amplication in all three sources, exhibiting narrow (roughly 1 km/s FWHM) emission features that are coincident (kinematically and/or spatially) with observed 22 GHz features. Moreover, in the case of W49N - for which observations were available at three epochs spanning a two year period - the spectra exhibited variability. The observed 621 GHz/22 GHz line ratios are consistent with a maser pumping model in which the population inversions arise from the combined effects of collisional excitation and spontaneous radiative decay, and the inferred physical conditions can plausibly arise in gas heated by either dissociative or non-dissociative shocks. The collisional excitation model also predicts that the 22 GHz population inversion will be quenched at higher densities than that of the 621 GHz transition, providing a natural explanation for the observational fact that 22 GHz maser emission appears to be a necessary but insufficient condition for 621 GHz maser emission.


Introduction
Barely fifteen years after the invention of the laboratory maser in 1953 -which required a population inversion in the ammonia molecule to be carefully engineered -interstellar masers were discovered as a remarkable and naturally-occurring phenomenon. The 6 16 − 5 23 transition of water vapor, with a frequency near 22 GHz, was among the first masing transitions detected from an astrophysical source (the Orion Molecular Cloud; Cheung et al. 1969), and has proven to be an extraordinarily useful probe in environments as diverse as interstellar shock waves, circumnuclear gas in active galactic nuclei (AGN), and the envelopes of evolved stars (Elitzur 1992;Lo 2005; and references therein). With brightness temperatures that often exceed 10 10 K and in extreme cases can exceed 10 15 K (e.g. Garay, Moran & Haschick (1989), and with linewidths that are often extremely narrow, 22 GHz water masers can be observed with Very Long Baseline Interferometry (VLBI). The very high angular resolution of such VLBI observations enables proper motion studies of the kinematics of gas outflowing from proto-and evolved stars. Moreover, trigonometric and geometric/rotational parallaxes can be determined, yielding the distances and motions of star forming regions in our Galaxy, in Local Group galaxies, and in more distant AGN (e.g. Brunthaler et al. 2007;Braatz et al. 2010). Such observations have led to revised estimates of the size, shape and kinematics of the Milky Way (e.g. Reid et al. 2009), as well as the best evidence yet obtained for the existence of supermassive black holes (e.g. Miyoshi et al. 1995). Masers directly probe extreme environments (of density and temperature), whose properties can often only be inferred indirectly from other lines.
The detection of multiple masing transitions has been crucial in elucidating the pumping mechanism responsible for the population inversions and the physical conditions in the masing gas. Understanding the latter, of course, is crucial to the astrophysical interpretation of maser observations. While the interpretation of the emission observed in a single line (e.g. the 22 GHz transition) is very geometry dependent, because the non-linear amplification of the radiation strongly favors those sight-lines that happen to possess the greatest velocity coherence, the interpretation of multiline observations is more robust and permits important constraints to be placed upon (1) the excitation mechanism and (2) the conditions of temperature and density in the emitting gas. The pattern of pure rotational maser transitions observed from interstellar sources appears to confirm an excitation model (e.g. Neufeld & Melnick 1991) in which collisional pumping, combined with spontaneous radiative decay, leads to the inversion of exactly those transitions that are observed to mase.
In evolved stars, however, the additional presence of the 437 GHz 7 53 − 6 60 maser transition suggests that collisional pumping is not the entire story and that radiative pumping by dust continuum radiation is also important.
Quantitative measurements of maser line ratios provide additional constraints. For example, the predicted dependence of a maser line ratio upon the temperature and density -5of the masing gas was used by Melnick et al. (1993) to derive a lower limit of 1000 K upon the gas temperature in several observed sources; this value argued against a model (Elitzur, Hollenbach & McKee 1989) in which the masers were excited in 400 K material behind a dissociative shock wave, and favored a model (Kaufman & Neufeld 1996) in which non-dissociative magnetohydrodynamic shocks were the source of the emission. The analysis makes the assumption that the maser beam angle is similar for the two transitions that are being compared.
Prior to the launch of the Herschel Space Observatory (Pilbratt et al. 2010), the observational data on pure rotational water masers had been limited to transitions of frequency less than 500 GHz, largely because of atmospheric absorption. The HIFI spectrometer on Herschel (de Grauuw et al. 2010), however, provides the opportunity of expanding the available data set by the addition of higher frequency transitions that promise to increase our leverage on the pumping mechanism and the physical conditions in the maser-emitting gas. To date, Herschel observations of oxygen-rich evolved stars have led to the detection of two additional water maser transitions: (1) the 5 32 − 4 41 transition at 620.701 GHz 1 (detected toward VY CMa by Harwit et al. 2010, andtoward W Hya, IK Tau andIRC+10011 by Justtanont et al. 2012); and (2) the 5 24 − 4 31 transition at 970.315 GHz (detected toward W Hya and IK Tau by Justtanont et al. 2012). To our knowledge, neither maser has previously been observed from interstellar gas. In this paper, we report the first detection of 621 GHz water maser emission from interstellar gas, obtained with Herschel/HIFI in mapping observations of the Orion-KL region. In addition, we report the detection of 621 GHz water maser emission obtained serendipitously in single pointings toward the W49N star-forming region and the Orion South molecular cloud.

Observations and data reduction
Our Herschel observations of Orion-KL were carried out on 2011 March 11 as part of the "Orion Small Maps" subprogram within the HEXOS Guaranteed Time Key Program (GTKP; P.I., E. Bergin). We used the "Heterodyne Instrument for the Infrared" (HIFI), in "on-the-fly mapping" (OTF) mode, to obtain a Nyquist-sampled map consisting of a 6 by 8 rectangular array of pointings spaced by ∼ 16 ′′ in R.A. and declination. The map center was located at offset (∆αcosδ, ∆δ) = (+1.5 ′′ , +10.5 ′′ ) relative to Orion-KL. (All offsets given in this paper are relative to 5h 35m 14.3s, −5d 22 ′ 33.7 ′′ (J2000), the position we adopt for Orion-KL.) The reference position for the OTF mapping observations, at offset (+841.5 ′′ , +870.5 ′′ ), was chosen to be devoid of known molecular emission. The observations were carried out in the upper sideband of mixer band 1b, using the WBS spectrometer, which provides an oversampled channel spacing of 0.5 MHz (0.27 km/s at a frequency of 621 GHz), roughly one-half the effective resolution. The absolute frequency calibration is accurate to 100 kHz (Roefsema et al. 2012). The beam size was 34 ′′ (HPBW), and the absolute pointing accuracy is 2 ′′ . Two concatenated Astronomical Observation Requests (AORs) were used to obtain two separate maps, with a small relative offset (6.6 ′′ ) chosen to make the center of the H polarization beam for one map coincident with the center of the V polarization beam for the other map. As discussed in §3, the goal of acquiring two separate maps in this manner was to obtain a measurement of any linear polarization in the maser feature. As summarized in GHz 6 16 − 5 23 transition, with the goal of determining the 621 GHz / 22 GHz line flux ratio as a constraint upon models for the maser emission mechanism. Here, we obtained a map, centered at offset (+3.1 ′′ , +3.9 ′′ ), consisting of a 9 by 9 square array of pointings spaced by ∼ 20 ′′ in R.A. and declination. The beam size was 41 ′′ (HPBW), and the frequency resolution was 6.1 kHz, corresponding to a velocity resolution of 0.082 km/s. The observations were performed using the K-band receiver at the prime focus of the 100 m telescope, with an observing time of 1.0 min per position. In calibrating the spectra, we applied corrections for the atmospheric attenuation and for the dependence of the telescope gain on the elevation. The calibration factor was determined by the observation of suitable calibration sources like NGC7027 and 3C286 (taking into account the significant linear polarization of the latter).
In addition to the mapping observations of Orion-KL that are the primary subject of this paper, we have also identified two additional Herschel/HIFI spectra that show narrow and/or time variable 621 GHz features suggestive of maser action. Our observations of high-mass star-forming region W49N were carried out at three separate epochs, as  Table 1. The data reduction methods that we adopted within the PRISMAS GTKP have been described, for example, by Neufeld et al. 2010. Our observations of the hot core in the Orion South molecular cloud were performed as part of a spectral line survey in the HEXOS GTKP. The data of present interest were acquired in a full spectral scan of band 1b, in -8which the 621 GHz transition was observed (in either the upper or the lower sideband) at 19 separate LO settings. The data reduction methods that we adopted for spectral scans within the HEXOS GTKP have been described by Bergin et al. (2010), and further details of the observations appear in Table 1.
Ancillary 22 GHz maser observations were performed using the Effelsberg 100 m telescope toward W49N and Orion S on 2012 Oct 1 and 2012 Sep 28, respectively. Because water masers are well-known to exhibit significant variability on timescales of months, our comparison of these non-contemporaneous 22 GHz and 621 GHz spectra must be interpreted with caution; indeed, as described in §3 below, time variability in the maser emission is readily inferred from a comparison of the W49N 621 GHz data acquired at the 3 epochs.  (Melnick et al. 2010). However, to the northwest of Orion-KL, close to the shocked material associated with Orion H 2 Peak 1 (Beckwith et al. 1978), a narrow emission feature is clearly present. This feature is most apparent in the four spectra obtained closest to offset (−15 ′′ , +43 ′′ ), the average of which is shown in Figure 4 (e.g.  Figure 4, the 22 GHz rest frequency is taken as 22.23508 GHz (Kukolich 1969), the average over all hyperfine components. The three strongest hyperfine components (for which ∆F = −1 and which are expected to account for more than 99.8% of the observed emission,) lie at frequency shifts of −36, −3 and +40 kHz relative to the mean 6 16 − 5 23 line frequency, corresponding to velocity shifts of 0.49, 0.04, and −0.55 km/s. Thus, even if departures from LTE lead to large changes in the hyperfine line ratios, the maximum resulting velocity shift is at most 0.5 km/s; hyperfine splitting of the higher-frequency 621 GHz transitions is completely negligible.

Pointed observations toward W49N and Orion South
The upper panel of Figure 5 shows the 22 GHz spectrum obtained toward W49N (red), along with the 621 GHz spectra obtained at three epochs spanning two years (see Table 1). Because the 621 GHz transition was not specifically targeted but rather observed serendipitously in observations with a different primary purpose, we did not perform near-contemporaneous 22 GHz observations. Thus, the 22 GHz spectrum shown in Figure  Four interferometric studies provide information about the spatial distribution of the 22 GHz water maser emission in W49N over a 30-year time period (Moran et al. 1973;Walker, Matsakis, & Garcia-Barreto 1982;Gwinn, Moran & Reid 1992;McGrath, Goss & De Pree 2004;performed in 1970-71, 1978-82, and 1998. They indicate that 99% of the 22 GHz maser flux is associated with the ultracompact HII regions designated G1 and G2 (Dreher et al. 1984), and originates within 5 ′′ of the beam center position adopted for our Herschel (and Effelsberg) observations. The remainder of the observed 22 GHz maser emission originates from a region ∼ 19 ′′ northeast of the beam center. Given these offsets, the Herschel beam size (34 ′′ ), and the absolute pointing accuracy (2 ′′ ), we find it very unlikely that the observed 621 GHz variability could be the -11result of pointing errors. The only way that pointing errors could falsely indicate variability is if a very strong 621 GHz maser emission feature were located in the far wing of the Herschel beam profile (where the beam response is a strong function of offset); such an emission feature would lie outside the region from which 22 GHz emission has been detected in the interferometric studies discussed above.
In the case of the 621 GHz spectra obtained toward W49N, the relative contribution of maser and non-maser emission is hard to disentangle, at least based on the data currently in hand. While it is tempting to interpret the spectra as a superposition of narrow maser features on top of a broad (∼ 40 km/s wide) plateau of thermal line emission, the 22 GHz spectrum has an entirely similar appearance. In the case of the 22 GHz transition, however, the expected contribution of thermal emission is negligible; indeed, the interferometric studies referenced above show that the broad pedestal apparent in single-dish observations is -in reality -a superposition of numerous compact and narrow emission features within the beam.
The lower panel of Figure 5 shows the spectra obtained toward Orion South in Orion S is not unique to water. Maser emission from hydroxyl radicals has been long been studied in W49N, but is believed to require different excitation conditions (e.g. Mader et al. 1975). In Orion South, Voronkov et al. (2005) have reported 6.7 GHz CH 3 OH emissions that are likely masing.

Spatially-averaged line ratios
In Table 2, we present the 22 GHz and 621 GHz photon luminosities for the three sources we have observed, each computed with the assumption that the emission is isotropic and for adopted distances of 414 pc (Orion-KL and Orion-S; Menten et al. 2007) and 11400 pc (W49N: Gwinn, Moran & Reid 1992). As we have noted in §3 above, however, particular velocity components that are clearly present in the 22 GHz spectra are sometimes unaccompanied by detectable 621 GHz line emission (though the converse is never true), suggesting that the requirements for strong 621 GHz maser amplification are more stringent than those for strong 22 GHz maser action. Thus, the source-averaged 621 GHz/22 GHz line ratios may be considerable smaller than those applying to specific 621 GHz emission features. Accordingly, we have also computed luminosities and line ratios for specific velocity ranges in which strong 621 GHz emission is observed from Orion-KL and Orion-S.
Furthermore, as will be discussed further in §4.1.2 below, our mapping observations of Orion-KL indicate that even when integrated over a narrow range of LSR velocities, the 621 GHz / 22 GHz line ratio varies spatially.

Spatial variations in Orion-KL
In Figure 6, we present channel maps for the 22 GHz line emissions detected from Orion-KL, with the location of the peak intensity marked by a red diamond, and the channel-averaged peak flux (in Jy/beam) listed in red near the top of each panel. The two strongest velocity components, at v LSR ∼ 7.5 km/s and v LSR ∼ 11.8 km/s show peak intensities near (+4 ′′ , −6 ′′ ) and (+20 ′′ , +14 ′′ ) respectively, positions separated by 26 ′′ , more than one-half the Effelberg HPBW (41 ′′ ). 3 Moreover, as shown in Figure 7  To test this hypothesis, we have carefully compared the 10 -13 km/s 22 GHz channel map with the beam profile for the Effelsberg telescope. The latter was measured at 22 GHz by observing the compact continuum source 3C84 at the prime focus. The beam response function is shown in Figure 8 (black contours, upper panel), superposed on red contours indicating the best fit Gaussian (with a HPBW of 41.03 ′′ ). In each case, contours are labeled as a percentage of the peak response. The beam profile is reasonably well-approximated by a Gaussian, with deviations less than ∼ 5% of the peak intensity.
The lower panel shows a grayscale representation of the residuals relative to the best-fit Gaussian, with gray levels separated by 1% of the peak intensity and the blue contour represents the result of that subtraction, which we will henceforth refer to as the "22 GHz residual" (i.e. the residual with respect to the intensity expected for a point source at offset [+15.7 ′′ ,+13.6 ′′ ]). Here, adjacent grayscale levels are separated by 1000 K km/s, corresponding to ∼ 5% of the peak 22 GHz line intensity integrated over the 10 -13 km/s velocity range. The spatial distribution of the "22 GHz residual" matches that of the 621 GHz line emission reasonably well, supporting the interpretation introduced at the of end the previous paragraph. Thus, the ratio of the luminosity of the 621 GHz line to that of the 22 GHz residual provides our best estimate of the true maser line ratio in the 621 GHz-emitting gas: 0.28 (Table 2). It is difficult to provide a quantitative estimate of the systematic uncertainty in the 22 GHz residual; the latter depends upon the reproducibility of the Effelsberg beam profile shown in Figure 8, an issue that is not presently constrained by observations. We note, however, that the peak flux in the residual is ∼ 20% of the peak flux in the 10 -13 km/s velocity range. By comparison, the maximum deviations from a

Comparison with maser excitation models
To help interpret the observed maser line ratios, we have updated the maser excitation models described by (Neufeld & Melnick 1991;hereafter NM91) to make use of recent The optical depths plotted in Figure 9 are the values along sightlines parallel to the direction of the velocity gradient (i.e. in the direction of minimum velocity coherence.) In plane-parallel geometry, the optical depth of a masing transition formally approaches −∞ as the inclination of the ray approaches 90 0 , although in reality the magnitude of the optical depth is limited by departures from plane-parallel geometry (such as curvature in a shocked molecular shell, for example). Nevertheless, the maximum optical depth in a typical interstellar region can easily exceed the Sobolev optical depth by an order of magnitude: thus, a Sobolev optical depth of 3 can easily produce a maser gain of e 30 . Gain factors of this magnitude will reduce the population inversion, leading to maser saturation.
We computed the maximum photon luminosity that can be obtained under conditions of saturation, with the use of the formalism discussed in NM91. In essence, this is the maximum rate of stimulated emission that can occur without eliminating the population inversion. Taking the ratio of these luminosities for the two transitions yields the results shown in Figure 10 Melnick et al. 1993), the 621 GHz transition does not provide strong -18constraints on the gas temperature; thus the observed 621 GHz/22 GHz line ratios are consistent with either 400 K gas behind dissociative J-type shocks (e.g. Elitzur, Hollenbach & McKee 1989), or the hotter gas that can arise behind non-dissociative shocks (Kaufman & Neufeld 1993). A search for maser emission in the 6 42 − 5 51 (471 GHz) and 6 43 − 5 50 (439 GHz) transitions of water could provide a stronger constraint on the gas temperature, as these transitions can only be pumped significantly at temperatures ∼ 900 K or higher (Melnick et al. 1993).

Summary
We  in its C configuration, as part of a project (AC443) that targeted seven sources. The total observing time for that project was 3 hours, of which 16 minutes was devoted to on-source observations of Orion-KL. The extragalactic radio source 0605-085 was used as a phase calibrator. In the absence of any non-variable radio source that could be used for absolute flux calibration, we established the flux density scale by assuming a flux density for 0605-085 of 1.7 Jy, the value measured in 1997 December by (Kovalev et al. 1999). The data were reduced with Astronomical Image Processing System (AIPS) software package. After a standard flux and phase calibration, we used the strongest maser emission feature -which has a flux of 1500 Jy and a v LSR of -4.6 km s −1 -for self calibration, and then applied the solutions to all the other channels. The FWHM primary beam of individual VLA antennas is ≈ 2 ′ . We imaged and cleaned the data by applying the natural weighting of the UV data, which resulted in a synthesized beam of size 1 ′′ .3 × 1 ′′ .0 (FWHM) at position angle -23 • . We produced a 2048 ×2048 maser image cube, with a cell size of 0 ′′ .1 × 0 ′′ .1. We then used the AIPS task SAD to detect maser emissions with peak intensity higher than 5 times the r.m.s. noise level. For line channels in the Orion-KL region that are not limited by dynamic range, the r.m.s. noise level was ≈ 0.12 Jy.
In Table 3, we list the parameters obtained for 339 maser emission spots, all selected to satisfy the following criteria: flux ≥ 1 Jy, signal noise ratio (SNR) ≥ 10, and dynamic range ≥ 0.01. (In other words, for channels with peak flux greater than 100 Jy, we used a flux cutoff of 0.01 × the peak flux). Figure 11 shows distribution of maser spots, with the color and symbol size representing LSR velocity and peak flux respectively. For comparision with the 621 GHz maser feature observed toward Orion-KL, we show in Figure 12 the 22 GHz maser emission integrated over the 10 -13 km s −1 velocity range. The letters A, B, C, D, E and F mark subregions from which the spectra presented in Figure 13 were extracted.
These spectra were obtained at the peak of each subregion, after Hanning smoothing the data with a kernel of FWHM 1 ′′ .       Figure 12, by Hanning smoothing the data in spatial coordinates (with a kernel of FWHM 1 ′′ ). Some spectra show ringing artifacts associated with the Gibbs phenomenon; these can occur when a strong maser feature is narrower than the channel width.